The Pilot Worked. That Was the Problem.

There is a slide I have seen in maybe thirty decks, across four industries, in fonts ranging from corporate-approved to quietly desperate. It always comes right after the good news. And the good news is real: the pilot worked. Cycle time down, error rate down, the site team standing in a photo they didn't expect to be in. Then the slide: Phase 2 — roll out to twelve sites by Q3.

That slide has buried more good work than any budget cut I have witnessed. Not because scaling is hard — although it is — but because the slide misunderstands what the pilot proved.

A pilot that works at one site is proof of a solution. It is not proof of a rollout plan. Those are two different claims, and the space between them is where the second site quietly becomes the place where the first site's success goes to die.

What was actually running at site one

Walk back through any pilot that worked and take an honest inventory of what was true there. Not what the handover documentation says — what was actually true, on the floor, on a Tuesday.

There was a person whose name was attached to the thing. Not a “process owner” in a RACI chart — a human being whose next role, whose standing in the building, whose pride was welded to this working. She chased the blockers before they appeared in any report. She apologised to nobody for interrupting the old routine.

There was slack. Somebody senior had quietly granted the team room to absorb the disruption — a protected afternoon, a deadline that bent, a target that looked the other way for a quarter. Nobody called it an investment. It was one.

There was attention. The pilot was being watched, which meant problems surfaced at the coffee machine and got fixed within hours — often by the vendor's best engineer, who took the call because pilots get the A-team. And there was permission: the pilot site was allowed to deviate from the standard process, because it was the deviation.

None of this is in the deck. Nobody is hiding it — it's invisible because it never felt like part of the solution. It felt like circumstances. But remove it, and the same tool, the same process, the same training deck produce a different result. The pilot ran in a greenhouse. The rollout plants the seedling in a parking lot and blames the seedling.

Here is the thing I’d rather you feel than take on faith. Flip the handover.

The misdiagnosis

When the rollout stalls, the explanations arrive on schedule: the other sites are resisting change. We need more communication. We need more training. Each one is half true, which is what makes them so durable — and each one points at the people instead of the conditions, which is what makes them so useless.

The other sites are not resisting the solution. They are the control group the pilot never had. Site two is the first honest test of the thing itself, stripped of its champion, its slack, its attention and its permission — and the test is coming back with a result nobody wants to read. The rollout didn’t fail. The rollout measured.

This is also why “let’s just pick our strongest site for the pilot” is the most expensive sentence in transformation work. The strongest site is the one where the invisible conditions already exist. You are not de-risking the rollout — you are designing an experiment that cannot fail, which means it cannot teach.

What the ones that landed did differently

The work of mine that is still running years after I left — an internal coach programme now measured in the hundreds, a method that has run since 2018 in countries I have never set foot in — did not scale by copying artifacts. Nobody shipped a binder. What travelled was ownership: at every new site, in every new cohort, a person with the standing, the reason and the room to make the thing theirs. The conditions the pilot enjoyed for free were deliberately rebuilt, locally, every time — not exported from headquarters.

That is the reframe, and it fits in one sentence: you don’t roll out a pilot — you regrow, at every site, the conditions the pilot got for free. The unit of rollout is not the tool. It is a person plus the circumstances that let the tool matter. Budget for that the way you budgeted for licences, and the artifacts will take care of themselves.

A small test you can feel

Before the next Phase 2 slide gets approved, ask the pilot team one question — in private, with amnesty, and not in writing: “What made this work that is not in the handover documentation?”

The length of that list is your rollout risk, item by item. Each entry is a design task for site two — not a training module, a condition to rebuild. And if the list comes back empty, don’t be reassured. Be worried. It means nobody has looked.

Trying to work out why a pilot that worked is dying in the rollout — or to design a rollout that won’t? That is exactly the kind of thing worth sketching.